to find more info... hmm, it's kinda hard, as far as i know, the block-boundary reset thing is not really documented much in pd's documentation. there have been several discussions about it on the mailing list though, so you could try searching the list archives with terms like 'phase, block, phasor' etc
as for why...there are audio processes that rely on having audio in blocks (FFT etc), and also cpu-wise i think there are probably savings from triggering a block at a time rather than sample by sample. ..but the main reason is that when miller designed pd, he designed it to work in blocks. i think some other softwares don't set blocks unless you instruct them to. a few of my friends won't use pd precisely because of the block restrictions. it seems like a bit of an arbitrary decision, and there are undoubtably pros and cons to both approaches, but pd is built around blocks, and will almost certainly stay that way.